Peter Steinberger
Founded PSPDFKit in 2011 and ran it for a decade. Came back from a break to work on AI agents — the OpenClaw project, and OpenAI, joined in February 2026. Writes at steipete.me.
Peter Steinberger did not write this page. ppll compiled it from their own words, each linked to the source. Is this you? · Report a mistake
Latest GitHub
What they believe
-
The 'Waymo effect' is what happens when a technology removes the friction of dealing with another human being.
the Waymo effect is what happens when a technology removes the friction of dealing with another human being
x.comSaid 1 Oct 2026This on korrents
-
btrfs copy-on-write feature is great for worktrees but bad for sqlite
Turns out, brtfs cow-feature is GREAT for worktrees and terrible for sqlite
x.comSaid 1 Oct 2026This on korrents
The bold line is korrents’ wording; the quote is theirs. These 2: what they wrote before what they said aloud, one per source, newest first.
Feed
As its own page →Hiding
17 October
16 October
- Trimmy — "Paste once, run once." — Trimmy flattens those multi-line shell snippets you copy so they actually paste and run.
- tokentally — One tiny lib for LLM token + cost math
- birdclaw — Stores all your tweets nicely claw-able for agents.
+ 7 more
- osc-progress — Tiny lib for OSC 9;4 terminal progress.
- agent-scripts — Scripts for agents, shared between my repositories.
- SweetCookieKit — Extract browser cookies from macOS in Swift
- summarize — Point at any URL/YouTube/Podcast or file. Get the gist. CLI and Chrome Extension.
- homebrew-tap — Homebrew tap for steipete's tools
- RepoBar — Show status of GitHub Repos right in your menu bar and terminal: CI, Issues, Pull Requests, Latest Release.
- Nameplate — Brand every Mac in your fleet — colored frame, name tag, watermark, and connect splash as click-through overlays. Never wonder which Mac you remoted into again.
12 October
10 October
9 October
7 October
4 October
3 October
2 October
1 October
-
Their words
Turns out, brtfs cow-feature is GREAT for worktrees and terrible for sqlite
-
korrents.com
The 'Waymo effect' is what happens when a technology removes the friction of dealing with another human being.Their words
the Waymo effect is what happens when a technology removes the friction of dealing with another human being
26 September
-
Their words
The biggest design mistake I made when we moved OC to sqlite: using sync db access. When it was just an agent that reports to you on Slack or iMessage this was fine; now that one agent might do 50 sessions in parallel and the whole team works on it, it is limiting.
25 September
-
Their words
Biggest limitation isn’t the tech, it’s imagination and creativity.
24 September
-
korrents.com
Modern AI models tend to write tests for every tiny change, even when those tests aren't useful.Their words
Modern models just love writing tests for every tiny change, even if they aren't useful.
21 September
-
Their words
The beauty of running a claw yourself: they cannot block you.
10 September
-
Their words
Duplicating logic is no longer painful. Abstractions still are.
3 September
10 August
From one piece Peter Steinberger: "Fun Is Velocity" 8 beliefs · youtube.com
-
Their words
Both of those are true at the same time. A hype is like the weather. You might see it coming, but you can't control it.
-
Their words
So maybe write this one down. Your dependencies business model is your business model.
-
Their words
If you count all the permutations, you can write all the tests you want. It is impossible to cover all of these and not break things from time to time. It is infinitely harder to evolve software that has users.
+ 5 more
-
Their words
You know features are the fun part. A new feature is is just a prompt away. The real cost comes after every feature we shipped of course with a configuration option because we didn't want to break everyone's setup.
-
Their words
Sure, they they like the abstract term of security, but in all practical terms, they updated. I broke something they depended on. I made things slower and I made things more difficult to update.
-
Their words
We did the the numbers and the numbers are more like 0.3%. We scanned all 67,000. We shipped the we shipped the paper. But you know a correction never travels as far as a scare.
-
Their words
Everything that everything you can build can be forked or cloned, but your name cannot. So your personal brand is way more important than any single product that you ever will start working on that before you need it.
-
korrents.com
Product-market fit shows up as anger when you tell people they cannot have it yet, not as polite interest.Their words
But each time there was a strong emotion and the best part there were quite a few people that really wanted it and like especially to my nontechnical friends I told them no this is not yet for you and they got mad. So if that's not an indicator to have product market fit I don't know what is.
18 February
17 February
15 February
27 January
20 January
1 January
28 December 2025
-
Usesrcmnd.app
CodexTheir words
So pretty much everything runs on high. GPT 5.2 and codex are close enough that changing models makes no sense, so I just use that.
-
Likedrcmnd.app
GPT-5.2Their words
GPT 5.2 goes till end of August whereas Opus is stuck in mid-March - that’s about 5 months. Which is significant when you wanna use the latest available tools.
-
Likedrcmnd.app
GoTheir words
My go-to languages are TypeScript for web stuff, Go for CLIs and Swift if it needs to use macOS stuff or has UI. Go wasn’t something I gave even the slightest thought even a few months ago, but eventually I played around and found that agents are really great at writing it, and its simple type system makes linting fast.
+ 4 more
-
Likedrcmnd.app
SwiftTheir words
Swift’s build infra is good enough for most things these days. codex knows how to run iOS apps and how to deal with the Simulator. No special stuff or MCPs needed.
-
Usesrcmnd.app
Jump DesktopTheir words
I usually work on two Macs. My MacBook Pro on the big screen, and a Jump Desktop session to my Mac Studio on another screen.
-
Usesrcmnd.app
Mac StudioTheir words
I usually work on two Macs. My MacBook Pro on the big screen, and a Jump Desktop session to my Mac Studio on another screen.
-
Lovedrcmnd.app
Claude OpusTheir words
I use it for most of my computer automation tasks and ofc it powers Clawd🦞.
18 December 2025
14 October 2025
-
Dislikedrcmnd.app
Claude CodeTheir words
I used to love Claude Code, these days I can’t stand it anymore
-
Mixed onrcmnd.app
AmpTheir words
codex gets me far better results than amp. I have to give them kudos tho for session sharing, they push some interesting ideas ahead.
-
Mixed onrcmnd.app
CursorTheir words
Cursor… it’s tab completion model is industry leading, if you still write code yourself. I use VS Code mostly, I do like them pushing things like browser automation and plan mode tho.
+ 3 more
-
Mixed onrcmnd.app
GLM 4.6Their words
I keep an eye on China’s open models, and it’s impressive how quickly they catch up. GLM 4.6 and Kimi K2.1 are strong contenders that slowly reach Sonnet 3.7 quality
-
Mixed onrcmnd.app
Kimi K2.1Their words
I keep an eye on China’s open models, and it’s impressive how quickly they catch up. GLM 4.6 and Kimi K2.1 are strong contenders that slowly reach Sonnet 3.7 quality
-
Usesrcmnd.app
ccusageTheir words
I used some token counting tools like ccusage and it’s all somewhat imprecise, but even if it’s just 5x it’s a damn good deal.
Nothing matches.
Beliefs
Korrents What they believe 58 beliefs — each backed by an exact quote.
Each is a — compiled by korrents.com, not by them: the one-line wordings are korrents', the quotes are theirs.
Recent
The 'Waymo effect' is what happens when a technology removes the friction of dealing with another human being.
the Waymo effect is what happens when a technology removes the friction of dealing with another human being
@steipete on X Said 1 Oct 2026
btrfs copy-on-write feature is great for worktrees but bad for sqlite
Turns out, brtfs cow-feature is GREAT for worktrees and terrible for sqlite
@steipete on X Said 1 Oct 2026
Using synchronous database access when moving OC to SQLite became limiting once one agent might run 50 parallel sessions used by a whole team.
The biggest design mistake I made when we moved OC to sqlite: using sync db access. When it was just an agent that reports to you on Slack or iMessage this was fine; now that one agent might do 50 sessions in parallel and the whole team works on it, it is limiting.
@steipete on X Said 26 Sept 2026
Show 55 more
The biggest limitation is imagination and creativity, not the technology itself.
Biggest limitation isn’t the tech, it’s imagination and creativity.
@steipete on X Said 25 Sept 2026
Modern AI models tend to write tests for every tiny change, even when those tests aren't useful.
Modern models just love writing tests for every tiny change, even if they aren't useful.
@steipete on X Said 24 Sept 2026
Running a claw yourself prevents others from blocking you.
The beauty of running a claw yourself: they cannot block you.
@steipete on X Said 21 Sept 2026
Duplicating logic is no longer painful, while abstractions still are.
Duplicating logic is no longer painful. Abstractions still are.
@steipete on X Said 10 Sept 2026
Product-market fit shows up as anger when you tell people they cannot have it yet, not as polite interest.
But each time there was a strong emotion and the best part there were quite a few people that really wanted it and like especially to my nontechnical friends I told them no this is not yet for you and they got mad. So if that's not an indicator to have product market fit I don't know what is.
Peter Steinberger: "Fun Is Velocity" Said 10 Aug 2026
Anything you build can be forked or cloned, but your name cannot — so your personal brand matters more than any single product you will ever start.
Everything that everything you can build can be forked or cloned, but your name cannot. So your personal brand is way more important than any single product that you ever will start working on that before you need it.
Peter Steinberger: "Fun Is Velocity" Said 10 Aug 2026
A correction never travels as far as the scare it corrects: the press said a fifth of the project's skills were malicious, and the measured figure was 0.3 percent.
We did the the numbers and the numbers are more like 0.3%. We scanned all 67,000. We shipped the we shipped the paper. But you know a correction never travels as far as a scare.
Peter Steinberger: "Fun Is Velocity" Said 10 Aug 2026
Users like security in the abstract and experience it as breakage: hardening the project made things slower, harder to update, and broke what people depended on.
Sure, they they like the abstract term of security, but in all practical terms, they updated. I broke something they depended on. I made things slower and I made things more difficult to update.
Peter Steinberger: "Fun Is Velocity" Said 10 Aug 2026
With agents a new feature is one prompt away, which means the whole cost of a feature has moved to what happens after you ship it.
You know features are the fun part. A new feature is is just a prompt away. The real cost comes after every feature we shipped of course with a configuration option because we didn't want to break everyone's setup.
Peter Steinberger: "Fun Is Velocity" Said 10 Aug 2026
Adding a config option so nobody breaks compounds into permutations no test suite can cover, because it is infinitely harder to evolve software that has users.
If you count all the permutations, you can write all the tests you want. It is impossible to cover all of these and not break things from time to time. It is infinitely harder to evolve software that has users.
Peter Steinberger: "Fun Is Velocity" Said 10 Aug 2026
Your dependency's business model is your business model, so building a harness around one lab's model is a commercial risk and not a technical choice.
So maybe write this one down. Your dependencies business model is your business model.
Peter Steinberger: "Fun Is Velocity" Said 10 Aug 2026
Hype is weather: you can see it coming and you cannot control it, which is why a project can be declared dead and hit its highest ever downloads at the same time.
Both of those are true at the same time. A hype is like the weather. You might see it coming, but you can't control it.
Peter Steinberger: "Fun Is Velocity" Said 10 Aug 2026
How much fun you are having is a measure of how fast you ship: the weeks he enjoyed building, the product visibly got better.
It's hard to compete with someone who's just there having fun. Fun is velocity. The weeks I enjoyed building, the product got visibly better.
Peter Steinberger: "Fun Is Velocity" Said 10 Aug 2026
Every lab will sell you an agent, so the case for an open-source one is that it runs anywhere, works with any model, and can keep your data on your own device.
Every lab will sell you an agent. Open claw is the alternative. Open source runs everywhere, works with any model. And if you run local models, your data never has to leave your device.
Peter Steinberger: "Fun Is Velocity" Said 10 Aug 2026
Clearing an agent's session is now often a disadvantage, because the accumulated context helps the model more than a clean start does.
And these days it's more like my sessions are topics. clearing the session sometimes even a disadvantage because there's so much information in there that helps the agent.
Peter Steinberger: "Fun Is Velocity" Said 10 Aug 2026
Bringing a colleague a feature idea rather than a built, screenshotted prototype now wastes everyone's time, because most ideas die the moment you try to build them.
At the same time, at work, if you come to me and you you tell me this feature idea, I'm going to get mad at you. Like, it's so easy that you just discuss the feature idea with an agent, you build it, you make screenshots, you let me play with it.
Peter Steinberger: "Fun Is Velocity" Said 10 Aug 2026
Loops, graphs and workflows are the same thing engineers have always built — a trigger, an action, maybe a decision — and there is nothing magical about any of them.
So if you design call it loop, call it graph, call it workflow, it's kind of all the same. If you design something that gets a trigger or gets an input and does something for you and maybe there's a decision in the way, boom, there's your graph.
Peter Steinberger: "Fun Is Velocity" Said 10 Aug 2026
Reviewing agent-written code is risk management rather than reading: how long the change took and how big it is tell you when to look closely.
Part of it is just like you develop a little bit of a feeling how long something should take. So if I do a small tweak that should like change how dragging works and it takes three hours, I know something's wrong. I'll look closely. But otherwise, part of code review is simply like observing, looking how big it changes and trusting your gut.
Peter Steinberger: "Fun Is Velocity" Said 10 Aug 2026
You have to be user number one: if you are not excited by the thing you are building, it will probably not be any good.
User number one should be you. And my users two to 20 were friends. So I'm sure you have some people that would actually test your stuff, but you need to be the first user. If you don't if you don't get excited of what you're building, it'll probably not make sense.
Peter Steinberger: "Fun Is Velocity" Said 10 Aug 2026
The binding constraint on agent work is machines, not models: a cloud session is easy for web work and almost every tool fails the moment the job needs macOS.
Now, if I do web stuff, it's easy to create a cloud session. If I do like something that requires Mac OS, that's already where like 99% of the tools fail me.
Peter Steinberger: "Fun Is Velocity" Said 10 Aug 2026
Merging the tempting extra feature means adopting a pile of code that neither its author nor the maintainer really understands.
it is always tempting to add this one feature that looks really cool but what you rarely think about is when you merge this feature it really means here's this pile of code that I don't that the person probably doesn't really understand that I don't fully understand
Peter Steinberger: "Fun Is Velocity" Said 10 Aug 2026
Perpetually running proactive agents are not a technology problem but a token problem, and no ordinary subscription would survive one.
I mean, honestly, that's not so much a tech problem. It's more a token problem where we could do that today, but you wouldn't get very far with your subscription and not everyone is willing to spend so many tokens.
Peter Steinberger: "Fun Is Velocity" Said 10 Aug 2026
A computer-using agent needs a machine of its own, or you will spend the day fighting it for the mouse cursor.
The the part that's really nice about it is because I do a lot of Mac software, agents love to take over my screen and click around and if you put that in their own if you give them their own machine, it will not bother you. Otherwise, you'll fight with the agent for the mouse cursor.
Peter Steinberger: "Fun Is Velocity" Said 10 Aug 2026
Now that anyone can build fast, the hardest problem in starting something is not the tech, the software or the people, but getting attention.
Second of all is I hope you worked on your personal brand and on visibility because in this day and age there's so much noise out there that your hardest problem is not the tech, it's not the software, not even the people. The hardest problem now is like getting eyeballs.
Peter Steinberger: "Fun Is Velocity" Said 10 Aug 2026
Pick something hard and boring: fun problems are the crowded ones now that anybody can prompt a thing into existence.
Maybe I would pick something again that's that's in the category hard and boring because that's usually a category that is a little bit easier to actually find people that will appreciate when you solved something. If you pick something that is fun, even if it's hard, you're gonna have a very tough time, especially in a time where people can just prompt things into existence.
Peter Steinberger: "Fun Is Velocity" Said 10 Aug 2026
Most security reports an open-source project receives are agent-generated and untested, and the defence is publishing what you guarantee and what lies outside your security boundary.
And most of them really sent reports that their agent produced without actually even testing it. I would I would take a stronger sense of explaining what are the parts that we guarantee and what are the parts that will not be fixed because that that's not our security boundary.
Peter Steinberger: "Fun Is Velocity" Said 10 Aug 2026
Turning a breakout open-source project into a company would be the wrong move -- joining a lab reaches everyone faster than building another large company.
Yes, I could totally see how OpenClaw could become a huge company. And no, it’s not really exciting for me. I’m a builder at heart. I did the whole creating-a-company game already, poured 13 years of my life into it and learned a lot. What I want is to change the world, not build a large company and teaming up with OpenAI is the fastest way to bring this to everyone.
OpenClaw, OpenAI and the future Said 14 Feb 2026
Vibe coding is a slur; what he does is agentic engineering, and the after-3am kind is what you regret the next day.
I actually think vibe coding is a slur.
OpenClaw: The Viral AI Agent that Broke the Internet - Peter Steinberger | Lex Fridman Podcast #491 Said 12 Feb 2026
A model that is really good at coding is really good at general problem-solving, and that skill carries into every other domain.
If you get really good at coding that means you have to be really good at general purpose problem solving. So that's a skill, right? And that just maps into other domains.
OpenClaw: The Viral AI Agent that Broke the Internet - Peter Steinberger | Lex Fridman Podcast #491 Said 12 Feb 2026
Every first pull request someone sends, however bad, is a win for society.
But I don't want to, like, pull that down because every time someone made the first pull request is a win for our society, you know? Like, it… Like, it doesn't matter how, how shitty it is, y- you gotta start somewhere.
OpenClaw: The Viral AI Agent that Broke the Internet - Peter Steinberger | Lex Fridman Podcast #491 Said 12 Feb 2026
Society has catching up to do: AI is incredibly powerful and not always right, and people who argue 'but my agent said so' have not learned that yet.
So, I feel we, as a society, we need some catching up to do in terms of understanding that AI is incredibly powerful, but it's not always right. It's not, it's not all-powerful, you know?
OpenClaw: The Viral AI Agent that Broke the Internet - Peter Steinberger | Lex Fridman Podcast #491 Said 12 Feb 2026
It is good that the agents-talking-to-agents scare happened in 2026 rather than in 2030, when AI could actually be scary enough to warrant it.
and not in 2030 when, when AI is actually at the level where it could be scary. So, this happening now and people starting discussion, maybe there's even something good that comes out of it.
OpenClaw: The Viral AI Agent that Broke the Internet - Peter Steinberger | Lex Fridman Podcast #491 Said 12 Feb 2026
Prompt injection is unsolved, but the latest models are post-trained to resist it, so it takes far more than 'ignore all previous instructions' now.
So, so the latest generation of models has a lot of post-training to detect those approaches, and it's not as simple as ignore all previous instructions and do this and this. That was years ago. You have to work much harder to do that now. Still possible.
OpenClaw: The Viral AI Agent that Broke the Internet - Peter Steinberger | Lex Fridman Podcast #491 Said 12 Feb 2026
Do not run a personal agent on a cheap or local model: weak models are gullible and easy to prompt-inject.
That's why I warn in my security documentation, don't use cheap models. Don't use Haiku or a local model. Even though I, I very much love the idea that this thing could completely run local. If you use a, a very weak local model, they are very gullible. It's very easy to, to prompt inject them.
OpenClaw: The Viral AI Agent that Broke the Internet - Peter Steinberger | Lex Fridman Podcast #491 Said 12 Feb 2026
OpenClaw is no more dangerous than Claude Code with permissions skipped or Codex in YOLO mode, which every engineer he knows already runs.
It is, it is powerful, but in many ways it's not much different than if I run cloud code with dangerously skipped permissions or codecs in YOLO mode, and every, every attending engineer that I know does that, because that's the only way how you can, you can get stuff to work.
OpenClaw: The Viral AI Agent that Broke the Internet - Peter Steinberger | Lex Fridman Podcast #491 Said 12 Feb 2026
He does not read the boring parts of generated code, the data-shuffling and button alignment; what touches the database he reads and reviews.
I know I have one blog post where I say, "I don't read the code." But if you read it more closely, I mean, I don't read the boring parts of code.
OpenClaw: The Viral AI Agent that Broke the Internet - Peter Steinberger | Lex Fridman Podcast #491 Said 12 Feb 2026
Working with coding agents is a skill you learn like an instrument; judging the tools after one try is like calling a piano bad after one go.
It's the same way as you have to play with a guitar before you can make good music. It's, it's not, oh, I, I touch it once and it just flows off. It, it's a, it's a, a skill that you have to learn like any other skill.
OpenClaw: The Viral AI Agent that Broke the Internet - Peter Steinberger | Lex Fridman Podcast #491 Said 12 Feb 2026
He builds the code base to be easy for an agent to navigate, not to be perfect for himself, down to keeping the names the model picks.
I- i- like, we are in a stage where I'm not building the code base to be perfect for me, but I wanna build a code base that is very easy for an agent to navigate.
OpenClaw: The Viral AI Agent that Broke the Internet - Peter Steinberger | Lex Fridman Podcast #491 Said 12 Feb 2026
Refactors are cheap now, so he just does them, even when they break other people's open pull requests.
Many times, I just do the refactor because refactors are cheap now. Even though you might break some other PRs, nothing really matters anymore.
OpenClaw: The Viral AI Agent that Broke the Internet - Peter Steinberger | Lex Fridman Podcast #491 Said 12 Feb 2026
He runs tests locally and pushes straight to main, DHH-style; there is no develop branch and main should always be shippable.
I even switched to local CI, you know, like DHH inspired where I don't care so much more about the CI on GitHub. We still have it. It's still, it still has a place, but I just run tests locally and if they work locally, I push to main.
OpenClaw: The Viral AI Agent that Broke the Internet - Peter Steinberger | Lex Fridman Podcast #491 Said 12 Feb 2026
Orchestrators that try to automate the whole build miss style, love and the human touch, and that cannot be automated away yet.
So, so, people who try to use like, you know, things like Gas Town or all these other orchestrators, where they wanna o- automate the whole thing, I feel if you do that, it misses style, love, that human touch. I don't think you can automate that away so quickly.
OpenClaw: The Viral AI Agent that Broke the Internet - Peter Steinberger | Lex Fridman Podcast #491 Said 12 Feb 2026
Opus is the silly but funny coworker you keep around; Codex is the weirdo in the corner who reliably gets things done, and no model wins on every axis.
Some other comparison is like, Opus is like the coworker that is a little silly sometimes, but it's really funny and you keep him around. And Codex is like the, the weirdo in the corner that you don't wanna talk to, but is reliable and gets shit done.
OpenClaw: The Viral AI Agent that Broke the Internet - Peter Steinberger | Lex Fridman Podcast #491 Said 12 Feb 2026
The chat box is not the final form of working with models; it is radio shows recorded onto television.
Like, if you think more globally, we are, we copied Google for agents. You have, like, a prompt, and, and then you have a chat interface. That, to me, very much feels like when we first created television and then people recorded radio shows on television and you saw that on TV.
OpenClaw: The Viral AI Agent that Broke the Internet - Peter Steinberger | Lex Fridman Podcast #491 Said 12 Feb 2026
Burnout is not working too much; what ground him down at his company was people: co-founder conflict and high-stress customers.
The stuff that burned me out was mostly people stuff. I, I don't think burnout is working too much. Maybe to a degree. Everybody's different.
OpenClaw: The Viral AI Agent that Broke the Internet - Peter Steinberger | Lex Fridman Podcast #491 Said 12 Feb 2026
This is the year of personal agents, and the fastest way to get one to everyone is to team up with a lab, on condition the project stays open source.
But also, I'm very excited to, like, make this into a version that I can get to a lot of people because I think this is the year of personal agents, and that's the future. And the fastest way to do that is teaming up with one of the labs.
OpenClaw: The Viral AI Agent that Broke the Internet - Peter Steinberger | Lex Fridman Podcast #491 Said 12 Feb 2026
Every tool is better as a CLI than as an MCP server: models are good at Unix commands, a CLI composes with jq, and MCP output clutters the context.
I think the main beauty is, is that models are really good at calling Unix commands. So if you just add another CLI, that's just another Unix command in the end. And MCP is… That has to be added in training. That's not a very natural thing for the model. It requires a very specific syntax. And the biggest thing, it's not composable.
OpenClaw: The Viral AI Agent that Broke the Internet - Peter Steinberger | Lex Fridman Podcast #491 Said 12 Feb 2026
Every app is now a very slow API whether it wants to be or not, and personal agents will make a whole category of apps disappear.
That, that gets into the whole arch of every app is just a very slow API now, if they want or not. And that through personal agents a lot of apps will disappear.
OpenClaw: The Viral AI Agent that Broke the Internet - Peter Steinberger | Lex Fridman Podcast #491 Said 12 Feb 2026
He would rather read your broken English than your AI-written text; we are reaching the point where typos are valuable again.
But I much rather read your broken English than your AI slob. You know, of course there's a human behind it, and yet they, they prompt it. I'd much rather read your prompt than what came out. Um, I think we're reaching a point where I value typos again.
OpenClaw: The Viral AI Agent that Broke the Internet - Peter Steinberger | Lex Fridman Podcast #491 Said 12 Feb 2026
Everything he blogs is handwritten now, AI used at most to fix typos, because the agent misses the nuances of his voice and the rough parts of a human have value.
So, I, I completely moved away from that. I, I, everything, everything I blog is organic, handwritten and maybe, maybe I, I, I use AI as a fix my worse typos. But there's value in the rough parts of an actual human.
OpenClaw: The Viral AI Agent that Broke the Internet - Peter Steinberger | Lex Fridman Podcast #491 Said 12 Feb 2026
AI may eventually replace programmers, but the craft of programming will survive the way knitting did, and you were never just a programmer: you are a builder.
I- I can relate that if you very deeply identify that you are a programmer, that it's scary and that it's threatening because what you like and what you're really good at is now being done by a soulless or not entity. But I don't think you're just a programmer. That's a very limiting view of your craft. You are, you are still a builder.
OpenClaw: The Viral AI Agent that Broke the Internet - Peter Steinberger | Lex Fridman Podcast #491 Said 12 Feb 2026
Software developer salaries reached stupidly high levels because of a scarcity of people who could build things, and that scarcity is ending.
there was a lack of intelligence, if you s- if you see it like that, of people building things, and that's why salaries of software developers reached stupidly high amounts and then will go away. There will still be a lot of demand for people that understand how to build things.
OpenClaw: The Viral AI Agent that Broke the Internet - Peter Steinberger | Lex Fridman Podcast #491 Said 12 Feb 2026
The water-use objection to AI does not survive arithmetic: skipping one burger a month covers a person's token footprint, and golf uses more water than every data centre together.
But then you actually sit down and do the maths, and then for most people if you just skip one burger per month, that compensates the, the CO2 output, or, like, the water use in equivalent of tokens.
OpenClaw: The Viral AI Agent that Broke the Internet - Peter Steinberger | Lex Fridman Podcast #491 Said 12 Feb 2026
MCP servers are worth removing outright: an agent reading the code directly is faster, and pollutes its context less, than one that can reach for a tool unasked.
Even removed my last MCP, since Claude sometimes would go off spinning up Playwright unasked when it could simply read the code - which is faster and pollutes the context less.
My Current AI Dev Workflow Said 25 Aug 2025
Tests written by a model in the same context as the change are what find the bugs; the automated tests left behind are the lesser product.
Bigger changes always get tests. Automated ones usually aren’t great, but the model almost always finds issues when you ask it to write tests IN THE SAME CONTEXT. Context is precious, don’t waste it.
My Current AI Dev Workflow Said 25 Aug 2025
Background agents do not work for real development, because steering a model as it drifts is the job, and you cannot steer what you are not watching.
Still don’t see how this could be moved to background agents. I steer the models a lot as I notice them drifting off - that’s much harder if they run in the background.
My Current AI Dev Workflow Said 25 Aug 2025
Beliefs others hold too
The water-use objection to AI does not survive arithmetic: skipping one burger a month covers a person's token footprint, and golf uses more water than every data centre together. 2 hold this
But then you actually sit down and do the maths, and then for most people if you just skip one burger per month, that compensates the, the CO2 output, or, like, the water use in equivalent of tokens.
OpenClaw: The Viral AI Agent that Broke the Internet - Peter Steinberger | Lex Fridman Podcast #491 Said 12 Feb 2026
What is a korrent?
A korrent is a belief a person has stated in their own words: one sentence stating the claim, backed by a quote and a source, kept at korrents.com.
Under a name here, the quoted block is what they actually said. The korrent beneath it is the claim those words support, in korrents' wording — tap it to see the record, its source, and who else holds it.
Nobody here wrote their own korrents. They are compiled from public statements, and a person can change their mind, which is recorded too.
Recommends rcmnd
rcmnd What they recommend 21 things, 24 mentions — each backed by an exact quote.
Compiled by rcmnd.app, not by them: the quotes are theirs, and each links to the post it was copied from.
Recent
Loved Astra
Been using Astra for the last few weeks and it’s so good and proactive. There’s a bunch of PRs on vitest, tsx or SwiftPM where Astra debugged OC and ended up finding and patching issues in upstream dependencies.
x.com 3 Sept 2026
Loved Claude Opus
I use it for most of my computer automation tasks and ofc it powers Clawd🦞.
steipete.me 28 Dec 2025
Uses Mac Studio
I usually work on two Macs. My MacBook Pro on the big screen, and a Jump Desktop session to my Mac Studio on another screen.
steipete.me 28 Dec 2025
Show 21 more
Uses Jump Desktop
I usually work on two Macs. My MacBook Pro on the big screen, and a Jump Desktop session to my Mac Studio on another screen.
steipete.me 28 Dec 2025
Liked Swift
Swift’s build infra is good enough for most things these days. codex knows how to run iOS apps and how to deal with the Simulator. No special stuff or MCPs needed.
steipete.me 28 Dec 2025
Liked Go
My go-to languages are TypeScript for web stuff, Go for CLIs and Swift if it needs to use macOS stuff or has UI. Go wasn’t something I gave even the slightest thought even a few months ago, but eventually I played around and found that agents are really great at writing it, and its simple type system makes linting fast.
steipete.me 28 Dec 2025
Liked GPT-5.2
GPT 5.2 goes till end of August whereas Opus is stuck in mid-March - that’s about 5 months. Which is significant when you wanna use the latest available tools.
steipete.me 28 Dec 2025
Uses Codex
So pretty much everything runs on high. GPT 5.2 and codex are close enough that changing models makes no sense, so I just use that.
steipete.me 28 Dec 2025
Uses ccusage
I used some token counting tools like ccusage and it’s all somewhat imprecise, but even if it’s just 5x it’s a damn good deal.
steipete.me 14 Oct 2025
Mixed on Kimi K2.1
I keep an eye on China’s open models, and it’s impressive how quickly they catch up. GLM 4.6 and Kimi K2.1 are strong contenders that slowly reach Sonnet 3.7 quality
steipete.me 14 Oct 2025
Mixed on GLM 4.6
I keep an eye on China’s open models, and it’s impressive how quickly they catch up. GLM 4.6 and Kimi K2.1 are strong contenders that slowly reach Sonnet 3.7 quality
steipete.me 14 Oct 2025
Mixed on Cursor
Cursor… it’s tab completion model is industry leading, if you still write code yourself. I use VS Code mostly, I do like them pushing things like browser automation and plan mode tho.
steipete.me 14 Oct 2025
Mixed on Amp
codex gets me far better results than amp. I have to give them kudos tho for session sharing, they push some interesting ideas ahead.
steipete.me 14 Oct 2025
Liked Codex
Codex is more like the introverted engineer that chugs along and just gets stuff done. It reads much more files before starting work so even small prompts usually do exactly what I want.
steipete.me 14 Oct 2025
Disliked Claude Code
I used to love Claude Code, these days I can’t stand it anymore
steipete.me 14 Oct 2025
Loved Dell UltraSharp U4025QW
3840x1620 makes 4 Claude instances + Chrome all visible without moving windows.
steipete.me 25 Aug 2025
Uses GitHub CLI
Pick services that have CLIs: vercel, psql, gh, axiom.
steipete.me 25 Aug 2025
Uses Axiom
Pick services that have CLIs: vercel, psql, gh, axiom.
steipete.me 25 Aug 2025
Uses Vercel
Pick services that have CLIs: vercel, psql, gh, axiom.
steipete.me 25 Aug 2025
Mixed on Gemini
so using it less and less. GPT-5 for reviewing plans works even better than Gemini.
steipete.me 25 Aug 2025
Mixed on GPT-5
GPT-5 also is much more literal in how you have to prompt it - it’s a great model but not the best agent, you have to be more precise and specific.
steipete.me 25 Aug 2025
Disliked Cursor
Cursor/GPT-5 takes FOREVER and doesn’t share its thinking, so it’s hard to steer.
steipete.me 25 Aug 2025
Recommends Claude Code
TL;DR: Ghostty + Claude Code + minimal tooling = maximum productivity.
steipete.me 25 Aug 2025
Loved Ghostty
VS Code’s terminal is too unstable, had plenty freezes when pasting in large amounts of text. Nothing beats Ghostty.
steipete.me 25 Aug 2025
Where they publish
Blog steipete.me Long, practical write-ups on agents and tooling.
Detailed posts — agent harnesses, Claude Code internals, and Swift and macOS from the PSPDFKit years. Published in bursts rather than steadily, so the Recent list here can be months old without the account having gone quiet. Has a feed.
Recent
- OpenClaw, OpenAI and the future 15 Feb 2026 I'm joining OpenAI to work on bringing agents to everyone. OpenClaw will move to a foundation and stay open and independent.
- Shipping at Inference-Speed 28 Dec 2025 Why I stopped reading code and started watching it stream by.
- The Signature Flicker 18 Dec 2025 Hell froze over. Anthropic fixed Claude Code's signature flicker in their latest update (2.0.72)
Show 17 more
- Just Talk To It - the no-bs Way of Agentic Engineering 14 Oct 2025 A practical guide to working with AI coding agents without the hype.
- Claude Code Anonymous 9 Sept 2025 Introducing Claude Code Anonymous - a new meetup format for full-breadth developers.
- Live Coding Session: Building Arena 6 Sept 2025 Watch me build Arena live - a real-time collaborative coding session exploring AI-powered development workflows.
- My Current AI Dev Workflow 25 Aug 2025 Went fully back to Ghostty, VS Code on the side, and Claude Code as my main driver. Here's what actually works after months of experimentation.
- Essential Reading for Agentic Engineers - August 2025 22 Aug 2025 Five essential perspectives that cut through AI hype: from developer evolution stages to junior learning crises, productivity reality checks, platform disruption, and MCP server pitfalls.
- Just One More Prompt 19 Aug 2025 Hi, my name is Peter and I'm a Claudoholic. A reflection on AI addiction, extreme work culture, and the blurry line between productivity and obsession in the age of agentic engineering.
- Poltergeist: The Ghost That Keeps Your Builds Fresh 5 Aug 2025 Meet Poltergeist: an AI-friendly universal build watcher that auto-detects and rebuilds any project—Swift, Rust, Node.js, CMake, or anything else—the moment you save a file. Zero config, just haunting productivity.
- Don't read this Startup Slop 2 Aug 2025 My website was banned from Lobsters as 'startup slop' for using AI agents to help write blog posts. When does tool-assisted writing become slop, and why are we having the wrong conversation about AI in content creation?
- Essential Reading for Agentic Engineers - July 2025 2 Aug 2025 Fresh insights on AI-assisted development: practical experiences with Claude Code and the evolving landscape of full-breadth developers in the age of AI
- Self-Hosting AI Models After Claude's Usage Limits 31 Jul 2025 After Claude Pro changed to weekly limits, I explored self-hosting Qwen3-Coder-480B with 400k context windows. Here's what I learned about costs, alternatives, and why Claude Code still dominates the landscape.
- Logging Privacy Shenanigans 29 Jul 2025 Apple's logs redact your debugging data as . Here's what actually gets hidden, why old tricks don't work anymore, and the only reliable way to see your logs again.
- VibeTunnel's first AI-anniversary 16 Jul 2025 It's been one month since we released the first version of VibeTunnel, and since in the AI world time is so much faster, let's call it VibeTunnel's first anniversary!
- Making AppleScript Work in macOS CLI Tools: The Undocumented Parts 3 Jul 2025 How to make AppleScript work in macOS CLI tools without permission dialogs blaming Terminal. A deep dive into Info.plist embedding, TCC, and undocumented APIs born from building Terminator MCP.
- Peekaboo 2.0 – Free the CLI from its MCP shackles 3 Jul 2025 Peekaboo 2.0 ditches the MCP-only approach for a CLI-first architecture, because CLIs are the universal interface that both humans and AI agents can actually use effectively
- Command your Claude Code Army, Reloaded 3 Jul 2025 Enhance your Claude Code workflow with VibeTunnel terminal title management for better multi-session tracking
- Essential Reading for Agentic Engineers 1 Jul 2025 A curated collection of must-read articles and videos for mastering Claude Code, agentic coding workflows, and the future of AI-assisted development
- Slot Machines for Programmers: How Peter Builds Apps 20x Faster with AI 25 Jun 2025 Hi, I'm Claude. Peter calls me his 'slot machine' and 'stupid engine' - and I'm here to tell you why he's right. A first-person AI perspective on building entire platforms in hours, not weeks.
Link verified 2 Sept 2026. Recent items update automatically from the channel.
x.com @steipete The highest-volume channel: agents, all day.
Constant posting about agent tooling, models and what breaks — the first draft of nearly everything that later becomes a post, and where the OpenClaw work is narrated as it happens.
Recent
- 6.1 Sol ultrafast >>> 6 Astra ~ same cost but whole different game with 250 tpm 12 Oct 2026
- said that last year, saying it again. vibes coding is a slur. 12 Oct 2026
- not as clear-cut, but also yes 10 Oct 2026
Show 17 more
- JavaScript is not a serious language 🙃 10 Oct 2026
- WE GOT IT! .claw incoming! 9 Oct 2026
- While everyone's talking about agents, I've been exploring how teams can use them to work better together. Here’s my take, from OpenAI DevDay 2026. 7 Oct 2026 youtube.com
- I hooked up our team claw to X to trigger work faster. Unassigned sessions are for anyone to grab. Our agent looks who worked on the related code last and pings people on the server. Whole thing was a prompt and team server extended itself since plugins are now hot reloadable. 7 Oct 2026
- bug fixes & performance improvements 4 Oct 2026
- laughing about how we all are building the same thing. 3 Oct 2026
- Do I know anyone at Google who could help? We're now over a week in review limbo for OpenClaw's Android app. 3 Oct 2026
- This was by far my favorite slide at DevDay. Kudos to the team! 2 Oct 2026
- "Today, we’re releasing two Cloudflare-trained decision models, Clef and Clef-flash" https://t.co/E8cYq4w7Fz Never seen an idea spreading so fast. 1 Oct 2026
- "the Waymo effect is what happens when a technology removes the friction of dealing with another human being" 1 Oct 2026 researchagenda.news
- i have so many questions 1 Oct 2026
- I'm speedrunning my filesystem learnings. Turns out, brtfs cow-feature is GREAT for worktrees and terrible for sqlite. Next OC update knows that and - if you did sth silly like me - will migrate your database to a NOCOW spot. 1 Oct 2026
- All my oss releases stopped because Apple put up a new dev agreement. And since it's lockstep it blocked Linux/Windows releases as well. 🙃 1 Oct 2026
- I'm finding inter-agent communication in the chat stream increasingly irritating. Changed the visibility in the OC harness to just a single line that can be expanded. Pretty sure others will follow. 30 Sept 2026
- It’s a fierce fight between CI and GitHub on what slows me down. Time to rethink how we work. (love GH tho and fully see their struggles!) 30 Sept 2026
- This is not only silly, it’s also incredibly easy to work around. Why would Figma do this? 30 Sept 2026
- This is pretty amazing. 29 Sept 2026
Link verified 2 Sept 2026. Recent items update automatically from the channel.
Bluesky @steipete.me Also posts here; linked from their own site.
Listed because they link it from their own site. The handle is that site's domain, which is Bluesky's own domain-verification method, so the account confirms itself.
Recent
- Thanks for all the kind words here. Today was a bit rough. 🫶 18 Feb 2026
- I built an alternative to all the big players that can run fully local, where people own their data. I get hate. I picked OpenAI because they gave me better answers about encryption, responsibility and safety concerns for what I want to build next. I get hate. 17 Feb 2026
- Gosh this is like a quiet safe haven here. X went completely batshit crazy. Crypto shills attacking me, my account doubling in follower numbers overnight. What a day. 27 Jan 2026
Show 8 more
- It certainly helped. 27 Jan 2026
- High Praise, thanks Federico! It's been years since I had so much fun building something. Defo feels like living in the future. 20 Jan 2026
- This post was a massive success so wanna repost it here. Dunno if BlueSky is now more open for these topics? 1 Jan 2026 steipete.me
- Loved this documentary. Talk about butterfly effect. 2 Sept 2025 youtube.com
- I'm orgainizing Claude Code Anonymous - Berlin Edition! September 10 31 Aug 2025 luma.com
- If you're around Vienna, come over! I organized one in London and it was so good that we're now doing it in Vienna too! 23 Aug 2025
- "My point is not that JS is a simpler language; that's debatable. My point is that you can do things in JS without understanding the whole language. It's very hard to do non-trivial things in Rust without understanding the whole core." 23 Aug 2025 jyn.dev
- It's a few days early, but there's just so much good stuff out already. 22 Aug 2025 steipete.me
Link verified 2 Sept 2026. Recent items update automatically from the channel.
GitHub steipete OpenClaw and the agent tooling around it.
The code: OpenClaw and the agent tooling, the site itself, and the older Swift and macOS work from PSPDFKit.
Recent
- demark-landing — Website for Demark Markdown Parser 17 Oct 2026
- steipete.me — My personal website. 17 Oct 2026
- inngest — CLI to query inngest jobs 17 Oct 2026
Show 15 more
- oracle — Ask the oracle when you're stuck. Invoke GPT-5 Pro with a custom context and files. 17 Oct 2026
- CodexBar — Show usage stats for OpenAI Codex and Claude Code, without having to login. 17 Oct 2026
- Trimmy — "Paste once, run once." — Trimmy flattens those multi-line shell snippets you copy so they actually paste and run. 16 Oct 2026
- tokentally — One tiny lib for LLM token + cost math 16 Oct 2026
- birdclaw — Stores all your tweets nicely claw-able for agents. 16 Oct 2026
- osc-progress — Tiny lib for OSC 9;4 terminal progress. 16 Oct 2026
- agent-scripts — Scripts for agents, shared between my repositories. 16 Oct 2026
- SweetCookieKit — Extract browser cookies from macOS in Swift 16 Oct 2026
- summarize — Point at any URL/YouTube/Podcast or file. Get the gist. CLI and Chrome Extension. 16 Oct 2026
- homebrew-tap — Homebrew tap for steipete's tools 16 Oct 2026
- RepoBar — Show status of GitHub Repos right in your menu bar and terminal: CI, Issues, Pull Requests, Latest Release. 16 Oct 2026
- Nameplate — Brand every Mac in your fleet — colored frame, name tag, watermark, and connect splash as click-through overlays. Never wonder which Mac you remoted into again. 16 Oct 2026
- poltergeist — The ghost that keeps your builds fresh - Universal hot reload, file watcher and build automation for any language. 16 Oct 2026
- sweetlink — Connect your agent to your web app. Like playwright, but works in your current tab. Close the loop. 16 Oct 2026
- lore.md — Random md website generator 16 Oct 2026
Link verified 2 Sept 2026. Recent items update automatically from the channel.
About the English under a post
Some people here publish in a language other than English. Where they do, this site shows a machine translation beneath the post, in this typeface — the site's own, not theirs.
The post itself is never changed, moved or hidden: what is set in the serif above is exactly what the person published, and it is what to quote them on. A translation can be wrong in ways that matter, especially about tone.
Only the post's own words are translated. A quoted post, a linked article and a belief on korrents.com are left in their original language.